Research map / Image generation / Fast diffusion sampling
DiffRGD: An Inference-Time Diffusion Guidance Through Riemannian Gradient Descent
Jia-Wei Liao, Li-Xuan Peng, Mei-Heng Yueh, Min Sun, Cheng-Fu Chou, Jun-Cheng Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use DiffRGD guidance instead of ADMMDiff | ADMMDiff · FFHQ 256 inpainting (1000 DDIM steps) | PSNR: 32.38 -> 34.04 | Improved |
| use DiffRGD guidance instead of DSG | DSG · FFHQ 256 super-resolution 4x (1000 DDIM steps) | PSNR: 26.71 -> 27.77 | Improved |
| use DiffRGD guidance instead of DSG | DSG · FFHQ 256 Gaussian deblurring (1000 DDIM steps) | PSNR: 27.45 -> 26.80 | Hurt |
| use DiffRGD guidance instead of ADMMDiff | ADMMDiff · FFHQ 256 motion deblurring (1000 DDIM steps) | PSNR: 27.26 -> 25.84 | Hurt |
| use DiffRGD guidance instead of ADMMDiff | ADMMDiff · FFHQ 256 inpainting (1000 DDIM steps) | FID: 29.52 -> 21.88 | Improved |
| use DiffRGD guidance instead of ADMMDiff | ADMMDiff · ImageNet 256 inpainting (100 DDIM steps) | PSNR: 27.97 -> 31.16 | Improved |
| reduce DDIM steps from 1000 to 100 for DPS | DPS · FFHQ 256 motion deblurring | PSNR: 24.47 -> 20.63 | Hurt |
| reduce DDIM steps from 1000 to 100 for DiffRGD | DiffRGD · FFHQ 256 Gaussian deblurring | PSNR: 26.80 -> 27.46 | Improved |
| use DiffRGD instead of DSG for segmentation guidance | DSG · CelebA-HQ segmentation-map guidance | mIoU: 0.750 -> 0.804 | Improved |
| use DiffRGD instead of DSG for sketch guidance | DSG · CelebA-HQ sketch guidance | Sketch-l2: 21.36 -> 19.48 | Improved |
| increase guidance strength for DSG from 0.05 to 0.3 (ablation) | DSG · CelebA-HQ segmentation-map guidance | FID: 95.69 -> 248.77 | Hurt |
| increase guidance strength for DiffRGD from 10 to 100 (ablation) | DiffRGD · CelebA-HQ segmentation-map guidance | mIoU: 0.74 -> 0.86 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FreeDoM — FreeDoM: Training-Free Energy-Guided Conditional Diffusion Model (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Plug-and-Play Diffusion Distillation (CVPR 2024 · similarity 0.90)
- EVODiff: Entropy-aware Variance Optimized Diffusion Inference (NeurIPS 2025 · similarity 0.89)
- Diffusion Models Without Attention (CVPR 2024 · similarity 0.89)
- Tortoise and Hare Guidance: Accelerating Diffusion Model Inference with Multirate Integration (NeurIPS 2025 · similarity 0.88)
- Diffusion State-Guided Projected Gradient for Inverse Problems (ICLR 2025 · similarity 0.88)
- Manifold Preserving Guided Diffusion (ICLR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
